Atmospheric Chemistry and Physics www.atmos-chem-phys.net 1680-7316 1680-7324 6 7 2006 10.5194/acp-6-1747-2006 http://www.atmos-chem-phys.net/6/1747/2006/ http://www.atmos-chem-phys.net/6/1747/2006/acp-6-1747-2006.html http://www.atmos-chem-phys.net/6/1747/2006/acp-6-1747-2006.pdf 1747 1770 2006-05-24 Inverse modelling of the spatial distribution of NO_x emissions on a continental scale using satellite data I. B. Konovalov M. Beekmann A. Richter J. P. Burrows Institute of Applied Physics, Russian Academy of Sciences, Nizhniy Novgorod, Russia Laboratoire Inter-Universitaire de Systèmes Atmosphériques, CNRS, Créteil, France Institute of Environmental Physics and Remote Sensing, IUP/IFE, University of Bremen, Bremen, Germany The recent important developments in satellite measurements of the composition of the lower atmosphere open the challenging perspective to use such measurements as independent information on sources and sinks of atmospheric pollutants. This study explores the possibility to improve estimates of gridded NO_x emissions used in a continental scale chemistry transport model (CTM), CHIMERE, by employing measurements performed by the GOME and SCIAMACHY instruments. We set-up an original inverse modelling scheme that not only enables a computationally efficient optimisation of the spatial distribution of seasonally averaged NO_x emissions (during summertime), but also allows estimating uncertainties in input data and a priori emissions. The key features of our method are (i) replacement of the CTM by a set of empirical models describing the relationships between tropospheric NO₂ columns and NO_x emissions with sufficient accuracy, (ii) combination of satellite data for tropospheric NO₂ columns with ground based measurements of near surface NO₂ concentrations, and (iii) evaluation of uncertainties in a posteriori emissions by means of a special Bayesian Monte-Carlo experiment which is based on random sampling of errors of both NO₂ columns and emission rates. We have estimated the uncertainty in a priori emissions based on the EMEP emission inventory to be about 1.9 (in terms of geometric standard deviation) and found the uncertainty in a posteriori emissions obtained from our inverse modelling scheme to be significantly lower (about 1.4). It is found also that a priori NO_x emission estimates are probable to be persistently biased in many regions of Western Europe, and that the use of a posteriori emissions in the CTM improves the agreement between the modelled and measured data.